并发控制算法是数据库系统保证事务执行正确且高效的重要手段，一直是数据库工业界和学术界研究的核心问题之一.将并发控制算法的基本思想归纳为“先定序后检验”，并基于该思想对现有各类并发控制算法进行了重新描述和分类总结.基于在开源内存型分布式事务测试床3TS上的实际对比实验，系统性地探究了各类算法的优缺点和适用场景，为面向内存数据库的并发控制算法的后续研究提供参考.

The concurrency control algorithm is a key component to guarantee the correctness and efficiency of executing transactions. Thus far, substantial effort has been devoted to proposing new concurrency controls algorithms in both database industry and academia. This study abstracts a common paradigm of the state-of-the-art and summarizes the core idea of concurrency control algorithms as “ordering-and-verifying”. Then, the existing concurrency control algorithms are re-presented following the ordering-and-verifying paradigm. Based on extensive experiments under an open-source memory-based distributed transaction testbed called 3TS, it is systematically demonstrated the advantages and disadvantages of the mainstream state-of-the-art concurrency control algorithms. Finally, the preferable application scenario is summarized for each algorithm and some valuable references are provided for the follow-up research of concurrency control algorithms used in in-memory databases.

国家重点研发计划（2020YFB2104100）；国家自然科学基金（61972403，61732014）；中央高校基本科研业务费专项资金（20XNLG22）；中国人民大学-腾讯联合实验室联合项目基金